Inward foreign direct investment and industrial restructuring: micro evidence the Slovenian firms growth model *

Transcription

1 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial... Zb. rad. Ekon. fak. Rij vol. 24 sv Preliminary communication UDC : : (497.4) Inward foreign direct investment and industrial restructuring: micro evidence the Slovenian firms growth model * Katja Zajc Kejžar 1, Andrej Kumar 2 Abstract We examine the impact of inward foreign direct investment (FDI) on the growth of local firms in terms of employment and total factor productivity (TFP) for the Slovenian manufacturing sector in the period. The theoretically predicted channels through which inward FDI affects the firm dynamics in a host country prove to be in general significant. First, there is evidence of the direct impact of foreign firms through so-called direct technology transfer as foreign-owned firms have higher growth of TFP compared to domestically-owned firms after controlling for other determinants. Secondly, the entry of foreign firms stimulates the reshuffling of the resources from less to more efficient local firms. The firm selection process is, namely, characterised by the least efficient firms experiencing a drop in their employment growth upon a foreign firm s entry. Thirdly, regarding the productivity spillover effects from foreign to local firms we provide indirect evidence that they mostly operate through vertical linkages rather than within the same industry. In general, it seems that not all firms are equally able to benefit from foreign firms presence and that absorptive capacity plays an important role. Key words: foreign direct investment, firm growth, productivity spillovers, firm selection process, industrial restructuring JEL classification: F21, F23, L11, L25, C23 * Received: ; Accepted: Assistant, PhD, University of Ljubljana, Faculty of Economics, Kardeljeva plošćad 17, 1000 Ljubljana, Slovenia. Scientific affiliation: International economics. Phone: , Fax: , katja.zajc@ef.uni-lj.si, Personal website: 2 Full professor, PhD, University of Ljubljana, Faculty of Economics, Kardeljeva plošćad 17, 1000 Ljubljana, Slovenia. Scientific affiliation: International economics. Phone: , Fax: , E- mail: andrej.kumar@ef.uni-lj.si, Personal website:

2 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial Zb. rad. Ekon. fak. Rij vol. 24 sv Introduction Several recent empirical studies provide evidence that within-industry reallocations from less to more productive firms and the exit/entry process contribute significantly to average productivity growth and constitute an important mode of industrial restructuring (see Olley and Pakes (1996), Roberts and Tybout (1996), Pavcnik (2002), Tybout (2001)). At the same time, it is widely recognised that investment liberalisation is one of the most important triggers of the industrial restructuring process. In the article we aim to provide a deeper understanding of the role of incoming FDI in shaping and restructuring of domestic industries in a small transition country, such is Slovenia. For this purpose the article assesses the impact of inward FDI on domestic firm growth taking into an account the probability of firm survival based on the panel of Slovenian manufacturing firms in the last decade. Namely, knowing how the entry and presence of foreign-owned firms affect growth of domestic firms and probability of domestic firms exiting is an important task in assessing the role of inward FDI in the industrial restructuring of a host country. As indicated in the theoretical literature (reviewed in, for example, Caves (1996), Aitken and Harrison (1999), Blomström and Kokko (1997, 1998)), there are two main opposing effects through which inward FDI can affect a domestic firm s growth and survival: competition effect and productivity spillover effects. The entry of foreign firm (either through exports or FDI) namely disturbs the existing equilibrium in the host country and increases the intensity of competition. The competition also intensifies if the foreign firm was exporting prior to establishing local production in the host-country market since, by avoiding export costs, the foreign firm s competitive position is improved. The competition effect of inward FDI (foreign firm entry/presence) has clear implications for the exit and growth of domestic firms. By increasing competition in the host country, a foreign firm s entry and presence may lead to the crowding out of local firms. The entry of a foreign firm with lower marginal costs draws demand from domestic firms as the foreign firm has an incentive to increase production relative to its domestic competitors. Domestic firms are then forced to cut production and the least productive ones even to exit the market. As emphasised by Görg and Strobl (2003), regardless of the cost structure the increased production of foreign rivals will generally lead to a reduction of the output price which will shrink the price-cost margin and increase the probability of the exit of domestic firms and reduce the prospects for their future growth. The competition effect may also operate in the factor markets, whereby foreign firms may also crowd out domestic rivals by increasing factor prices in the economy. However, some local rivals might react to this intensified competition by investing in product and/or process upgrading. On the other hand, foreign firms activity in the host country may confer positive externalities on domestic firms which may result in a decrease in the unit costs of

3 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial... Zb. rad. Ekon. fak. Rij vol. 24 sv local firms and consequently in an increase in their output. 3 However, several studies assert that local firms must have an ability to learn and take advantage of the technology employed in multinational enterprises (MNEs). This so-called absorptive capacity is often related to the skill intensity and learning experiences of the firm. A domestically-owned firm might benefit from the presence of foreign firms through the several potential channels: (i) through the backward and forward linkages (customer-supplier links) between MNEs and domestic firms, (ii) MNEs training of local employees, (iii) demonstration effects, and (iv) competition from MNEs. Productivity spillovers may thus occur within the same industry (intra-industry spillovers), in vertically upstream and downstream related industries (inter-industry spillovers), or as a result of agglomeration (see Blomström and Kokko, (1997, 1998)). Since productivity is one of the key determinants of whether or not a firm exits and of the prospects for its growth, through spillover effects FDI also indirectly positively affects the domestic firm s survival and its growth performance. The nature of these effects and the relative importance of the channels depend on a number of FDI characteristics. Therefore, the net effect of these two offsetting effects cannot be generalised and hence need to be tested empirically. As pointed out by Kosová (2004) to distinguish and estimate both effects separately two different measures of foreign firm entry and presence should be included in empirical model specification. While there is a substantial body of empirical literature on spillover effects, evidence of the importance and mechanisms of competition effects is relatively rare. Moreover, most studies test the presence of productivity spillovers by estimating the production function on the level of firms or plants without controlling for the fact that FDI might have a significant impact on the exit of domestic firms. Since firms that are forced to exit the market due to increased foreign competition are less likely to benefit from the presence of an MNE these studies tend to overestimate productivity spillover effects. Several authors therefore argue that FDI s impacts and, in particular, the competition/selection effect are better analysed from the perspectives of industry and firm dynamics. Accordingly, that is the framework we have chosen for our empirical analysis. More specifically, we estimate firm growth model on firmlevel data in Slovenian manufacturing sector in the period 4. The firm dynamics framework allow us to asses and characterize competition effect and the role of inward FDI in industry selection process more directly through its impact on the crowding out of local firms (increased probability of exiting and reduced growth). 3 These externalities may appear since technology is to some extent a public good, there is a belief These externalities may appear since technology is to some extent a public good, there is a belief that foreign firms may not be able to fully internalise their technological advantages and therefore their presence would lead to various types of productivity spillovers to domestic firms. 4 The reason for selecting the period for the empirical analysis was the availability of data. The reason for selecting the period for the empirical analysis was the availability of data. The data for the year 2003 were the latest available at the time of the article formation.

4 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial Zb. rad. Ekon. fak. Rij vol. 24 sv At the same time this framework allows us to indirectly test also for the productivity spillover effects. Despite the relatively numerous empirical studies on a firm s growth and survival the empirical evidence of the impacts of inward FDI or even of foreign competition generally on the growth and survival of local firms is very limited. The rare exceptions are studies by Görg and Strobl (2003) for Ireland, De Backer and Sleuwaegen (2003) for Belgium, and Kosová (2004) for the Czech Republic. Görg and Strobl (2003) investigate whether the presence of multinational companies measured as an MNE s share of employment in an industry has any effect on the survival of plants in the same sector using plant-level data for the Irish manufacturing sector in the period and employing a Cox proportional hazard model. They find that the presence of multinationals has a positive effect on the survival of Irish plants but this effect is significant only for plants that operate in high-tech sectors. Regarding foreign-owned plants, they provide evidence that foreign plants have higher hazards of exiting than indigenous plants and that the presence of multinationals has a negative effect on the survival of other foreign-owned plants in low-tech sectors. On the contrary, De Backer and Sleuwaegen (2003) find evidence that inward FDI increases domestic exits by separately estimating exit and entry functions for Belgian manufacturing data. The crowding-out effect is stronger in the case of FDI than in the case of imports. Moreover, they find that increased foreign competition both through increased imports and inward FDI negatively affects the entry of domestic firms. However, the empirical results suggest that the importance of positive longterm structural effects measured by the relative number of foreign firms in related industries (defined as industries belonging to the same NACE-2 digit level) between foreign and domestic firms can moderate or even reverse crowding-out effects. Kosová (2004) test the presence of static and dynamic crowding-out effects of domestic firms by foreign ones and technological spillover effects in survival and growth models using firm-level panel data for the Czech Republic. She finds no evidence that foreign firms are expanding in Czech markets at the expense of domestic firms or that foreign expansion induces the excessive exit of domestic firms. In contrast, she provides evidence that foreign expansion, measured by the foreign sales growth rate, has a positive effect on both the growth and survival of domestic firms. These results, together with significantly higher exit rates of domestic firms around the time of foreign entry, suggest that crowding-out, and thus the adjustment of domestic firms to FDI inflows, is merely a one-time static effect realised upon foreign entry. So there is a shakeout of domestic firms when foreign firms enter but, subsequent to this initial entry effect, domestic firms benefit from the expanding foreign industry.

5 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial... Zb. rad. Ekon. fak. Rij vol. 24 sv The rest of the article is organised as follows. In Section 2, the data and main descriptive statistics with respect to domestic- and foreign-owned firms are presented. Subsequent Section 3 specifies growth model to be applied to panel data on the Slovenian manufacturing firms and define the variables. Section 4 discusses main econometric problems dealt with in our empirical analysis. Further, Section 5 presents the results and implications of the empirical estimations. Finally, Section 7 concludes with the summary of the main findings of the empirical analysis. 2. The data We analyse the firms growth performance based on annual panel data on firms operating in the Slovenian manufacturing sector (NACE 15-37) in the period 5. One of the advantages of this dataset is that, unlike in most studies on FDI that restrict the sample to larger firms, our data cover the whole population of manufacturing firms as reporting is mandatory for all firms (business entities) registered in Slovenia. Originally, the dataset contained information on 9,711 firms operating between 1994 and 2003 but firms with a zero number of employees and a negative value of equity were dropped from our sample which gives 7652 firms in our final sample. The main descriptive statistics and the indicators of the importance of foreign firms in Slovenian manufacturing sector are presented in Tables 1 and 2. Foreign-owned firms are defined as firms in which foreign owners have at least 10% equity share. Table 1: Average size, labour and total factor productivity, real average annual wage and export propensity of domestic and foreign firms in Slovenian manufacturing sector, Employment Labour productivity TFP Annual wage Capitalintensity Export propensity 1000 SIT (1994 prices) dom. firm in y. 1994= SIT (1994 prices) 1000 SIT (1994 prices) Dom For Dom For Dom For Dom For Dom For Dom For Financial data were obtained from the database of firms financial statements collected by the Agency Financial data were obtained from the database of firms financial statements collected by the Agency of the Republic of Slovenia for Public Legal Records and Related Services. The data on firms formation, legal and organisational forms and termination of operation were obtained from the Business Register of Slovenia. Other data were provided by the Statistical office of Republic of Slovenia.

6 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial Zb. rad. Ekon. fak. Rij vol. 24 sv Source: Own calculations Note: Summary statistics exclude firms with 0 employees and non-positive equity reported. According to Table 1 there is a notable and persistent difference between domestic and foreign firms with respect to all characteristics seen in Table 1. As expected, throughout the whole period foreign firms demonstrated a higher average size, higher labour and total factor productivity (TFP) and higher capital intensity compared to domestic firms with the only exception of capital intensity in Foreign firms also pay higher average wages and are more export-oriented than their domestic rivals, selling around half of their output abroad. A slight convergence between domestic and foreign firms can be seen in terms of average labour productivity, which for domestic firms increases by approximately 80% versus the 60% increase in the average labour productivity of foreign firms within the period. On the other hand, domestic firms have been unable to decrease the gap in total factor productivity. The average total factor productivity of domestic firms remains practically unchanged throughout the period, while foreign firms face an approximate 30% increase compared to the initial year The data suggest that if the constant returns to scale is an acceptable assumption then the value-added increase per employee (labour productivity) of domestic firms can be almost exclusively attributed to the improvement in the capital equipment of labour (on the assumption of constant returns to scale), while improvements in total factor productivity play an important role in foreign firms. In Table 2 we report some indicators of the importance of foreign firms in the Slovenian manufacturing sector to motivate our expectations regarding the impacts of foreign firm entry and their presence on the performance of domestic firms. As can be seen in Table 2, the importance of foreign-owned firms increased during the period considered according to all measures as a share in the number of firms, employment, fixed assets and value added. At the end of our period, foreign firms accounted for approximately 18% of manufacturing employment and above 20% of both fixed assets and value added.

7 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial... Zb. rad. Ekon. fak. Rij vol. 24 sv Table 2: The relative importance of foreign firms in Slovenian manufacturing sector according to the selected indicators, No of firms Empl Fixed assets value added Domestic Foreign for. firm share for. firm share for. firm share for. firm share Source: Own calculations Note: Summary statistics exclude firms with 0 employees. The above-average performance of foreign firms satisfy the principal starting point of theories on MNEs, postulating that to be able to compensate for the inherent disadvantages of operating in foreign markets MNEs have to possess certain capabilities which give them some sort of competitive advantage over their domestic rivals. The question open to empirical testing is then whether foreign firms with their superior performance have a significant impact on the performance of local firms in the way the theory suggests. 3. Empirical model specification Both the theoretical and empirical literature often studies firm growth and exits together as an outcome of a single economic process of industrial evolution. Our growth model specification follows Evans (1987) approach where growth is modelled as a function of initial size:, (1) where S t denotes the size of the firm. Among the set of factors that affect a firm s growth we include various factors proposed by the different theoretical and empirical studies.

8 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial Zb. rad. Ekon. fak. Rij vol. 24 sv Most of the recent studies adopt the framework of so-called firm and industry dynamics models which focus on the selection process among heterogeneous firms within a particular industry that operates through the entry and exit process and emphasise the importance of firms learning process for the selection and evolution process within the industry. These models are thus also known as learning models as the entrant typically does not know its own cost structure (efficiency), but its relative efficiency is discovered through the processes of passive (Jovanovic, 1982) or active learning (Erikson and Pakes, 1995) from actual market experience subsequent to entry. The factors proposed by these theories can be classified into the following groups: (i) firm characteristics (λ t ); (ii) industry or product market characteristics (γ ), (iii) t factor prices (τ t ), and other exogenous factors (θ t ) that reflect conditions outside the domestic industry (τ t and θ t factors are both captured by the inclusion of annual dummies). Through a logarithmic transformation of (1) we obtain the firm-growth rate equation, (2) where u t is a normally distributed error term with a mean zero. Two versions of the firm growth model (2) are estimated: (i) the extensive growth model version where the firm s size is measured by its employment (empl t ) which gives firm s growth defined as the difference in the log values of the firm s number of employees (lnempl t+1 - lnempl t ), and (ii) the intensive growth model where firm s growth is measured in terms of TFP growth (lntfp ijt+1 -lntfp ijt ). For the functional form of G( ), we follow Evans (1987) approach and test a higherorder logarithmic expansion in two principal firm-specific variables (firm s size and age) until there is no evidence of further nonlinearity. Similarly as in several other studies third-order logarithmic expansion in firm s size, second-order logarithmic expansion in firm s age and a first-order logarithmic expansion in other variables was confirmed which yields the following regression equation of firm growth model: growth ijt+1 = b 0 +b 1 MNEentry jt +b 2 MNEentry dtfplow jt +b 3 dtfplow ijt + b 4 entry jt + b 5 hfdi jt +b 6 hfdi lnwage ijt +b 7 ReghFDI jrt +b 8 BackFDI jt + b 9 BackConc jt +b 10 ForFDI jt +b 11 ForConc jt +b 12 fdi ijt +b 13 HHI jt +b 14 plants ij + b 15 dexport it +b 16 dexmajor it +b 17 dprofit ijt + b 18 lnempl ijt +b 19 lnempl2 ijt + b 20 lnempl3 ijt +b 21 lnkint ijt +b 22 lntfp ijt +b 23 lnwage ijt +b 24 lnage ijt + b 25 lnage2 ijt +Σb 26,t dyear t +Σb 27,j dindustry j +Σb 28,r dregion r + Σb 29,o downertype l +h i +ε ijt (3)

9 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial... Zb. rad. Ekon. fak. Rij vol. 24 sv where subscripts i, j and t refer to firms, industries and years, respectively. growth ijt+1 denotes either firm s growth in terms of employment defined as the difference in the log values of the firm s employment (lnempl ijt+1 -lnempl ijt ) or firm s growth in terms of TFP (lntfp ijt+1 -lntfp ijt ). Ln in variable names denotes the natural logarithm of a particular variable, while 2 (sq) denotes that the variable enters the estimation in a squared form and 3 (cube) that the variable enters the model in its third power. All values of the financial variables are deflated using producer prices indices at the 2- digit NACE classification. Firm characteristics Among the principal firm characteristics that affect firm s growth the theories postulate firm s size, age and productivity. The size of a firm (empl ijt ) is measured by the number of employees. Age ijt denotes a firm s age counting from the formation year according to the Business Register. As age enters our empirical models in a logarithmic form we start to count age with a value of 1 in order to prevent the dropping of observations in the first year of firm s operation, which would generate sample selection bias due to the relatively high infant mortality rates. We also test the robustness of the results by including an age variable in non-logarithmic form and the results are robust. Productivity is measured as total factor productivity (TFP ijt ) based on production function estimates. Firm dynamics models predict that smaller and younger firms grow faster and are less likely to survive than old and large firms. This predicted size-growth relationship sharply contradicts the Gibrat s traditional law of independence between the growth of a firm and its size. The productivity of the firm is expected to negatively affect the likelihood of exit and positively affects the firm s growth. Further, we include capital-intensity Kint, measured by real fixed assets per worker. ijt The capital intensity of a firm is expected to positively affect its ability to survive and grow. According to the Olley and Pakes (1996) model, the stock of physical capital affects the distribution of future plant productivity 6. In this case, capital intensity may act as a proxy for other unobserved sources of efficiency leading to the higher likelihood of an exit and lower growth for low-capital-intensity plants. Wage ijt is defined as the average yearly real wage per employee. Unfortunately, data on the skill structure of employees is not available; therefore we also use the wage variable as a proxy for the skill intensity of a firm. This implies we are assuming that wages for similar education level/qualification categories of workers are similar across firms and industries. The real wage is also used as a proxy for human capital in Mata and Portugal (2004). Skill intensity is expected to positively affect a firm s 6 There is a relationship between a producer s underlying efficiency and the incentive to invest in There is a relationship between a producer s underlying efficiency and the incentive to invest in capital. Essentially, efficient firms generate higher levels of investment and larger capital stocks.

10 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial Zb. rad. Ekon. fak. Rij vol. 24 sv survival ability and its growth potential as it can serve as a proxy for its absorptive and learning capacity, which is a key determinant of the course of the firm s life. As we use a firm as a unit of observation we must control for the number of the firm i s subsidiaries (plants ij ) as the theory suggests that hazard and growth rates differ between a single-and multi-plant firm. Among the firms characteristics we additionally control for a firm s profitability by including the dummy variable dprofit ijt which equals 1 for firms with a positive net profit in year t. For testing the impact of the exporting we include two additional dummy variables: dexport it and dexmajor it are dummy variables for exporters. dexport it equals 1 for all exporters (positive sales in foreign markets), while dexmajor it takes the value of 1 in the case that export propensity the share of a firm s output supplied to foreign markets in the firm s total sales is greater than 70%. Industry characteristics Besides the time-invariant market characteristics that are captured in the set of industry dummies, we include the Herfindahl-Hirschman index HHI jt to measure market concentration. HHI jt is defined as the sum of the squares of the market shares of all firms within a particular industry at the 5-digit NACE level. The market share of firm i is defined as the share of its domestic market sales in total industry sales in the domestic market (all firms local sales + imports in industry j). The expected effect of market concentration it is not so clear-cut. On one hand, the concentration ratio is expected to have a positive impact on the survival and growth of firms. The argument is that the price level is more likely to be elevated above the long-run average cost at the minimum efficient scale level of output in concentrated industries which may facilitate the survival of suboptimal scale firms which is what typical entrant firms are. On the other hand, firms in highly concentrated markets may be subjected to fierce aggressive behaviour by rivals which may reduce their chances of survival. Our principal explanatory variables refer to the entry and presence of foreign owned firms to test for both direct and indirect effects of inward FDI measured by the extent of foreign affiliates operations in the host country. Concerning the indirect impacts of inward FDI on domestic firm dynamics, we test for the presence and relative strength of two opposing effects as predicted by theory: the competition effect and productivity spillovers. To distinguish between these two effects two different measures of foreign firm entry and their performance are included in the regression model. The competition effect is tested with the entry rate variables MNEentry jt and entry jt at the 3-digit level of the NACE classification. MNEentry jt denotes the foreign firm entry rate defined as the number of foreign entrants (greenfield and acquisitions) divided by the total number of firms operating in the industry j, and entry jt as an entry rate considering all entrants including domestic and foreign ones. Both measures exclude the firm for which the observation is taken. MNEentry jt tests the crowd-

11 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial... Zb. rad. Ekon. fak. Rij vol. 24 sv ing-out effect which takes place upon foreign firms entry, while entry jt serves as a controlling variable to control for the impact of a new firm entry in general. It is possible that all firm entries in a particular industry rather than foreign firm entries alone affect the exit decision and growth of incumbent firms so we therefore want to control for this possibility. As it is quite likely that MNEs are attracted to industries that offer favourable conditions which also stimulates domestic firm entry levels, estimates that do not control for this possibility may lead to an overestimation of the crowding-out effect of foreign firms entry. MNEentry dtfplow jt is the interaction term between MNEentry jt and dtfplow ijt, where dtfplow ijt is a dummy variable equal to 1 for firms in the lowest quintile in terms of total factor productivity at 3-digit level of NACE and 0 for other more efficient ones. If the least efficient firms are more likely to be crowded out by the entry of a foreign firm this interaction term should have a negative effect on firm growth and survival (a negative sign in the growth equation). The presence of horizontal (intra-industry) spillover effects is tested by the variable hfdi jt that measures the concentration of foreign firms in industry j as the foreign firms share in total industry employment: ; (4) where n denotes the number of all firms in industry j and empl ijt denotes the number of employees in firm i. fdi ijt is a dummy variable for foreign ownership. It takes a value of 1 for foreign firms considering a 10% ownership share threshold. The measure excludes the firm for which the observation is taken. The employment share of foreign firms is used in many studies testing the presence of horizontal spillover effects, among others (Barrios et al., 2005), (Keller and Yeaple, 2003), (Görg and Strobl, 2003). Instead of the employment share, other studies consider also foreign firms share in the industry s output (Smarzynska, 2004), and the relative number of foreign firms (De Backer and Sleuwaegen, 2003). Some studies also take into account the share of foreign equity participation in foreign firms, including (Aitken et al., 1999) and (Smarzynska, 2004). hfdi lnwage ijt is the interaction term between hfdi jt and ln(wage ijt ). As described above, we use ln(wage ijt ) as a proxy for a firm s absorptive capacity. If firms with a higher absorptive capacity are more able to take advantage of a foreign firm s presence in the industry, this term should have a positive sign in the growth equation and thus provide some support for the validity of our hypothesis that intra-industry spillover effects are stronger for local firms with a higher absorptive capacity (skill intensity).

12 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial Zb. rad. Ekon. fak. Rij vol. 24 sv ReghFDI jrt measures regional intra-industry foreign firm concentration in terms of employment share and tests whether any intra-industry spillovers are reinforced when domestically-owned firms are located close to foreign firms. More specifically, it is defined as:, (5) where m denotes the number of firms within industry i and region r. We consider regions at the NUTS 3 level. The inter-industry spillover effects are tested through two additional explanatory variables BackFDI jt and ForFDI jt measuring the concentration of FDI in backwardly- and forwardly-linked industries with industry j. BackFDI jt measures the extent of potential contacts between local suppliers and foreign firms (vertical connections between local suppliers and foreign affiliates customers) and thus tests the presence of backward inter-industry spillovers:, (6) where technical coefficient α jk denotes the share of product j originating from domestic production that is used by industry k in its intermediate consumption (excluding final use and imports of intermediate products). This variable accounts for the impact of foreign affiliates on their upstream local suppliers, that is for the impact of the concentration of foreign firms in industries to which industry j supplies its output. Following Smarzynska (2004), inputs supplied within an industry are not included since this effect is accounted for by the variable measuring the horizontal spillovers - hfdi jt. The extent of potential contacts between local customers and foreign firms-suppliers (through forward linkages) is measured by ForFDI jt :, (7) where the technical coefficient δ jk denotes the share of input k in the total intermediate consumption of industry j. 7 This variable accounts for the impact of foreign affili- 7 Technical coefficients α Technical coefficients α jk and δ jk are obtained from the input-output table, more specifically from Use table for the domestic output at basic prices. As the input-output table for the Slovenian economy is not available for all years in our sample, the year 2000 s I-O table was chosen as a base for the technical (input) coefficient calculation. BackFDI jt and ForFDI jt are constructed at the two-digit level of NACE which is the most detailed level of the I-O table available.

13 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial... Zb. rad. Ekon. fak. Rij vol. 24 sv ates on their downstream local customers (the impact of the concentration of foreign firms in industries that provide inputs for industry j). A positive and statistically significant coefficient in the growth model would suggest there are indeed positive inter-industry externalities connected to the concentration of foreign-owned firms in vertically linked industries. To control for the possibility that the general concentration of economic activity in interrelated industries rather than the concentration of foreign firm activity alone positively affects growth of local firms and that at the same time FDI is attracted to the prosperous industries, we include BackConc jt and ForConc jt as controlling variables and thus avoid any potential overestimation of vertical spillover effects. BackConc jt and ForConc jt are defined as backwardly- and forwardly-linked industries share of total manufacturing employment weighted by technical coefficients a and δ. To control for the industry-, time- and region-specific effects throughout our sample period we include annual dummies dyear t, industry dummies at the 3-digit level of NACE dindustry j, region dummies at the NUTS level dregion r, and dummies for ownership type downertype l discriminating among different types of ownership. We test several other variables but due to insignificant coefficients in all empirical specifications we do not include them in our final empirical models. Among others, we test for the effect of the ratio of long-term debt to total assets, for the minimum efficient scale defined as the log of median employment size in industry j and industry growth with respect to the previous year defined as the growth of total employment within particular industry j. 4. Econometric issues We estimate different empirical specifications of employment growth model (3) for the aggregate sample of domestic and foreign firms and for the sub-sample of domestically-owned firms separately. To test additional aspects of FDI impact on performance of the local firms we estimate also TFP growth equation. There are several potential econometric problems of estimating growth models. How we deal with most problematic ones is discussed below. Since we use panel data the estimation techniques allow us to control for firm-specific effects that are constant over time and are not explicitly represented in the model. Two standard panel data models, fixed effects models (FEM) and random effects models (REM), will be estimated. As Hausman test rejects the null hypothesis of no correlation between the right hand side variables and the random effects in our growth models, we report the results of fixed effects estimations on a sample of surviving firms (observation in the year of exit is treated as a missing value). For ro-

14 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial Zb. rad. Ekon. fak. Rij vol. 24 sv bustness purposes we estimate TFP growth model also by Prais-Winsten regression, where disturbances are assumed to be panel-level heteroskedastic with first-order autocorrelation AR(1) 8 within panels. Selection bias A potential concern in our model specification is the sample selection bias that generally refers to the situation where the dependent variable is not observed for a restricted, non-random sample. In the growth model it appears since small firms that have slow or negative growth are more likely to exit and thus disappear from the sample than are larger firms. The sample selection bias has long been recognised in growth models and properly dealt in most studies since Evans (1987) and Hall (1987). We correct for the sample selection bias by standard Heckman selection model in which we control for the probability of survival. This procedure consists of two equations, a firm exit equation used as a selection equation in the first step and a firm growth equation in the second stage where Mill s inverse ratio enters as an additional regressor to control for the probability of a selection into the sample of firms for which growth is observed. Endogeneity Another potential econometric concern which may cause biased estimates when testing for the direct impact of foreign ownership on a firm s growth is the possibility that the foreign ownership dummy variable (fdi ijt ) might not be entirely exogenous. It is usually argued that foreign investors tend to acquire shares in the most successful and larger firms and that foreign ownership is thus not randomly distributed (see Djankov and Hoekman (2000), Evenett and Voicu (2001), Damijan et al. (2003)). In this case, fdi ijt is potentially a choice variable that might be correlated with unobservables relegated to the error term. More specifically, fdi ijt may be endogenous if the decision for FDI (in the form of foreign acquisitions) is correlated with unobservables that affect a firm s exit decision/growth. For instance, if foreign investors are more likely to acquire shares in more successful firms and therefore experience lower probability of exiting or higher growth ceteris paribus, then if we fail to control for this correlation our results will underestimate (overestimate) the effect of foreign ownership on the probability of exiting (growth). To deal with this problem we instrument for fdi ijt and employ instrumental variables models. We use a two-stage least squares fixed-effects (within) and a random-effects estimator for the growth model. The instruments employed are size, size squared, age, age squared, the ratio of net profits to sales, export propensity, total factor productivity and average wage. We use lagged values of these instruments for domestic 8 Arellano-Bond test confirms that average autocovariance in residuals of order 1 is present.

15 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial... Zb. rad. Ekon. fak. Rij vol. 24 sv firms, values of the instruments in the year before an acquisition takes place for firms that have been acquired by foreign investors and the first-year values for greenfield FDI. To avoid autocorrelation we drop the first-year observations for greenfield investments. Total factor productivity estimates There are additional potential econometric concerns related to the estimation of the firm s total factor productivity. Typically, total factor productivity is estimated as the residual in the production function estimates based on firm-level panel data. Simultaneity bias is usually referred to as the endogeneity of production inputs, caused by a correlation between unobservable productivity shocks and input levels causing the regressors and the error term to be correlated which makes OLS estimates inconsistent. Bias thus occurs when at least part of the TFP is observed by the firm early enough to allow the firm to change its factor input decision. Several methods of controlling for simultaneity bias are proposed in the literature. Olley and Pakes (1996) developed an estimator that uses investment as a proxy for these unobservable productivity shocks. One of the drawbacks of Olley and Pakes (1996) approach is that there must be a strictly monotonous relationship between the proxy (investment) and output for obtaining the consistent estimates. This means that observations with a zero investment have to be dropped from the sample. Further, Levinsohn and Petrin (2003) point out that an investment is associated with substantial adjustment costs which make the investment very lumpy and not respond smoothly to a productivity shock, thus violating the consistency condition. Therefore, Levinsohn and Petrin (2003) develop a similar two-step estimator which uses intermediate inputs as proxies, arguing that intermediates may respond more smoothly to productivity shocks and may respond more fully to the entire productivity term than investment. We follow this approach in the estimation of TFP. Using intermediate input proxies instead of investment also allows us to avoid truncating observations with a zero investment Firm employment growth 5. Empirical results In Tables 3 and 4 we report the results for the employment growth model using various model specifications. Table 3 displays the results of fixed effects estimations considering only surviving firms (observation in the year of exit is treated as a missing value) for the aggregate sample of domestic and foreign firms (columns 1-3) and for the sub-sample of domestically-owned firms (column 4), whereas Table 4 represents the results from Heckman selection model in which we control for the

16 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial Zb. rad. Ekon. fak. Rij vol. 24 sv probability of survival. To control for the possible enodogeneity of foreign ownership status in the estimates based on aggregate sample, we employ two-stage least squares fixed-effects (within) estimator (column 3, Table 3). Based on all empirical specifications of employment growth model (Tables 3-4), the coefficient of MNEentry is negative but insignificant suggesting that in general foreign firm entry doesn t have significant impact on the employment growth of local firms within the same industry. The expectation that a foreign firm s entry does not affect the employment growth equally for all surviving firms is tested by the inclusion of the interaction term MNEentry dtfplow jt (columns 2-4, Table 3). It turns out that this interaction term has a negative and highly significant impact, while MNEentry jt remains insignificant but positive which indicates that there is crowding out effect upon foreign firm entry but it is only significant for the least efficient firms from the lowest quintile in terms of total factor productivity, while there is no significant evidence that more efficient firms would suffer a negative impact on their growth. This result lends support to our prediction that the least efficient local firms experience the downsizing after a foreign firm entry. The effect of intra-industry foreign firm concentration on local firms growth (intraindustry spillovers) tested with hfdi is on average insignificant. But again after inclusion of the interaction term with real average wage variable (lnwage) that is used as a proxy for the absorptive capacity both coefficients of hfdi jt and hfdi lnwage jt become significant at 1% (column 2-4 in Table 3) supporting the intuition that firms are not equally able to benefit from the presence of MNEs within the same industry. It is confirmed that negative impact of horizontal MNE concentration on firm s employment growth is decreasing in firm s skill-intensity suggesting that the probability of net positive impact of foreign firm presence on the employment growth of local surviving firms within the same industry is higher for skill-intensive firms. However, there is no evidence that intra-industry effects would be intensified in the case of the local concentration of foreign firm activity where local is defined by being located in the same region (insignificant coefficient of the ReghFDI jt ). Based on the fixed effects estimates considering only surviving firms (Table 3), the coefficients of variables that measure concentration of foreign firms activity in vertically linked industries (BackFDI and ForFDI) are significant in all specifications, but with the opposite signs. The concentration of foreign firms activity positively affects the growth in their upstream local suppliers (through backward linkages). This result is in line with recent findings of positive productivity externalities connected to the extent of foreign firm presence through backward linkages for several transition countries on the sample of larger surviving firms (for instance, Damijan et al. (2003) for Czech Republic, Poland and Slovenia, and Smarzynska (2004) for Lithuania).

17 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial... Zb. rad. Ekon. fak. Rij vol. 24 sv Table 3: Fixed effects estimates of firm growth model for surviving firms FE FE Instrumental variab. FE All firms All firms All firms Domestic firms MNEentry (-0.49) (0.35) (0.75) (0.33) MNEentry dtfplow -0.6** (-2) *** (-2.67) ** (-2.23) dtfplow -0.03*** (-3.52) *** (-2.65) *** (-3.36) entry (0.4) (0.33) (-0.03) (0.82) hfdi (-1.48) -0.73*** (-3) *** (-3.02) *** (-2.62) hfdi lnwage 0.096*** (2.84) 0.117*** (2.88) 0.086** (2.44) ReghFDI (1) (0.98) (1.24) 0.134* (1.75) BackFDI 0.391** (2.07) 0.362* (1.9) 0.404** (2.05) 0.344* (1.72) BackConc 3.66** (2.09) 3.307* (1.81) 3.074* (1.64) 3.906** (2.01) ForFDI ** (-2.4) -0.36* (-1.86) * (-1.81) ** (-2) ForConc 1.53 (0.96) 1.78 (1.09) (0.18) (0.73) HHI (0.94) (0.84) (0.32) (0.61) dexport 0.014** (2.02) 0.014** (2.03) 0.015** (2.11) 0.014** (1.98) dexmajor 0.042*** (3.72) 0.043*** (3.73) 0.049*** (4.11) 0.050*** (4.11) dprofit 0.078*** (9.19) 0.077*** (9.05) 0.076*** (8.75) 0.080*** (9.06) fdi INS (0.63) (0.53) (-0.51) lnempl *** (-49.4) -0.59*** (-49.5) *** (-44.11) FE *** (-48.54) lnempl *** (11.6) 0.066*** (11.6) 0.08*** (12.9) 0.069*** (11.35) lnempl *** (-6.13) -0*** (-6.1) *** (-7.99) *** (-5.91) lntfp 0.04*** (5.71) 0.023*** (2.98) 0.04*** (4.57) 0.023*** (2.76) lnkint 0.04*** (14.4) 0.039*** (14.2) 0.044*** (14.57) 0.038*** (13.25) lnwage 0.12*** (18.8) 0.108*** (14.8) 0.147*** (16.21) 0.104*** (13.85) lnage *** (-11.7) -0.25*** (-11.7) *** (-2.65) *** (-11.23) lnage *** (8.25) 0.134*** (8.4) 0.074*** (2.59) 0.140*** (8.42) R F test that a u_i=0 1.55*** (0.000) 1.55*** (0.000) 1.54*** (0.000) 1.5*** (0.000) time dummies INCL INCL INCL INCL N Source: Own calculations Notes: - t-statistics are in parentheses, - INS denotes instrumented, - ***, **, * denotes significance at 1%, 5% and 10%, respectively.

18 Katja Zajc Kejžar, Andrej Kumar Inward foreign direct investment and industrial Zb. rad. Ekon. fak. Rij vol. 24 sv However, after correcting for the probability to exit (Table 4), the impact of foreign firm activity through backward linkages is no longer significant. The results suggest that some firms face an increased probability of exiting in response to the concentration of foreign firms in downstream industries, while the successful surviving firms are able to take advantage of the MNEs presence in downstream industries through backward linkages. This leads us to suspect that the degree of linkages might be weaker in the case of MNEs compared to domestic firms, leading to a decrease in demand for intermediate products. As pointed out by Smarzynska (2004) and Saggi (2002), MNEs entry to downstream sectors might lower the demand for domestically produced intermediates, particularly when it forces less productive domestic producers in this industry to exit. The demand for domestically produced intermediates might decrease either because they use inputs more efficiently or they rely more on imported intermediates (see Rivera-Batiz and Rivera-Batiz (1991)). In addition, some local suppliers might not be able to achieve the higher product standards or delivery conditions demanded by foreign firms. Table 4: Heckman selection model of firm growth ALL FIRMS DOMESTIC FIRMS 1 2 plants (0.78) (0.89) MNEentry (-0.49) (-0.5) entry (0.19) (0.22) hfdi (-0.87) (-0.7) ReghFDI (0.88) (1.37) BackFDI (1.45) (1.08) BackConc (0.87) (1.67) ForFDI -0.43** (-1.96) *** (-2.39) ForConc 3.199* (1.82) (1.49) HHI (-0.41) (-0.82) dexport (0.19) (0.4) dexmajor 0.037*** (4.79) 0.039*** (4.88) dprofit 0.092*** (10.14) 0.093*** (10.57) fdi (-1.45) / lnempl -0.8*** (-68.8) *** (-72.29) lnempl *** (25.34) 0.140*** (27.45) lnempl3-0.01*** (-17.8) *** (-19.75) lntfp 0.021*** (2.64) 0.024*** (3.09) lnkint 0.028*** (8.39) 0.027*** (8.3) lnwage 0.078*** (10.51) 0.075*** (10.4) lnage -0.05*** (-2.48) *** (-2.73) lnage2-0.01* (-1.71) (-1.51) Cons *** (-2.61)